would be really useful if you can run the debug executable, but not by double clicking, instead running it in a cmd. this way, when it crashes, you can see output in console. this will be nearly impossible to fix without an error log
that shoudl be fine
open cmd, cd to the folder of your f95checker install, then just type F95Checker-Debug.exe
thats normal, the fucker writing the opengl library made it where it just prints that with no option to disable it, and it doesnt matter to normal usage
depends whats happening. if its an exception, there should be a traceback. if the sorting issues are related to the exception causing the crash, this will help. if its a kernel level process crash, like a segmentation fault, then might be less useful but hopefully should still give some hint
Ah, good to hear, I was just about to come back to post it happened to me in 1301.
ive seen it corrupt the sorting 4 times in the past hour right after a refresh. its seemingly random. just decides to corrupt the memory where imgui stored sorting info after certain actions.
Manual refreshing or auto-refresh in the system-tray? I never use auto-refresh and only manual, but also with hours in between each next refresh, maybe that's why I haven't seen it happen then?
No, it was manual refreshing. Sometimes it's switching tab, sometimes making installed, sometimes changing labels, something archiving, sometimes changing executables. Seems like anything that triggers a resort/refilter has a chance to cause a memory corruption to the sort specs
1 good piece of news atleast, the timeout happened again tonight and it recovered successfully. so it was in fact just a bug with aiohttp that had already been fixed upstream. i have now also added proper parsing of timeout errors, so they are considered "normal" (as in just a connection issue to be retried shortly, not an error, so now if a timeout like this happens while updating cache for a certain game, it will be marked with error f95zone unavailable and be retried after 15 mins ( as opposed to the generic exception's 6 hour retry timer)
Is this the message you refer to?
No. That's client side. Rate limit is very different than a timeout. Timeout means no response within the designated time, rate limit means the server responded that you are asking too much and need to slow down. The timeout refers to the issue that was happening on my server, the rate limit you see is happening between your pc and the forum. You requests to f95zone are already slowed down to 1 every 2 seconds, and when a rate limit response is received it retries up to 10 times. So when you see this message is because you're rate limited and it will not get better soon. It has already tried for 10 times over 20 seconds, continuing to make the request hoping that the rate limit passes is pointless. And that's just for getting the notifications, which is just 2 requests: check if logged in, request notification counts. And you're getting fully rate limited for over 20 seconds. That just shows how aggressive the rate limit is and how not using the new cache api was impossible moving forward.
Ok, but I thought that the new cache api would prevent this from happening. e.g. the checker would not be accessing the forum at all during a refresh apart from checking for notifications or when doing a quick search.
that is precisely what i said. the ratelimit popup you got is due to being ratelimited from simply checking the notifications. disable checking for notifications, and you will not get any ratelimits. and again, you got that popup because simply checking the notifications was rate limited for 10 times in a row over 20 seconds.
possibly ran out of memory, seems like the only possible explanation to me. drawing images with opengl is memory intensive, and browsers are huge memory hogs too. when you run out of memory and dont have swap space the kernel just kills programs at random i believe
on another note, i may have found a way to reduce the impact, where notifications are less expensive to check (no strict need for xftoken), and quick search doesnt require login (can use latest updates api).
then im sorry, i have no clue what is going wrong or any way to find out more. if more users report these random crashes i will try to look into it.
and search has more options as a nice bonus of this change:
